A Fully-Decentralized Consensus-Based ADMM Approach for DC-OPF With Demand Response

TLDR

Smart‑grid complexity from distributed data and emerging techniques drives the need for distributed optimization, prompting the use of a consensus‑based ADMM approach to solve dynamic DC‑OPF with demand response. The study aims to quantify how key parameters and subsystem partitioning affect convergence and data traffic in distributed DC‑OPF, and to provide guidance verified on large‑scale systems. The authors implement a fully decentralized ADMM that solves local DC‑OPF problems in parallel, coordinating subsystems through boundary‑bus phase‑angle consensus, and propose an accelerated variant to improve convergence while limiting communication to adjacent subsystems. All three algorithms reliably satisfy network, generator, and demand‑response constraints and guarantee global convergence, and the study offers validated guidance for subsystem partitioning on large‑scale systems.

Abstract

This paper discusses a consensus-based alternating direction method of multipliers (ADMMs) approach for solving the dynamic dc optimal power flow (DC-OPF) problem with demand response in a distributed manner. In smart grid, emerging techniques together with distributed nature of data and information, significantly increase the complexity of power systems operation and stimulate the needs for distributed optimization. In this paper, the distributed DC-OPF approach solves local OPF problems of individual subsystems in parallel, which are coordinated via global consensus variables (i.e., phase angles on boundary buses of adjacent subsystems). Three distributed DC-OPF algorithms are discussed with different convergence performance and/or communication requirement. All three distributed algorithms can effectively handle prevailing constraints for the transmission network, generating units, and demand response in individual subsystems, while the global convergence can be guaranteed. In particular, based on the traditional distributed ADMM approach, a fully decentralized algorithm without the central controller is proposed in Algorithm 2 with a new communication strategy, in which only limited information on boundary buses are exchanged among adjacent subsystems. In addition, the accelerated ADMM is discussed in Algorithm 3 for improving the convergence performance. In recognizing distributed OPF approaches in literature, one major research focus on this paper is to quantify the impact of key parameters and subsystem partitioning strategies on the convergence performance and the data traffic via numerical case studies. A general guidance for subsystem partitioning is proposed and verified via large-scale power systems.